Survival after a diagnosis of cancer is affected by a variety of individual, tumour and healthcare system factors. Individual factors include sex, age at diagnosis, comorbidity, socioeconomic status and lifestyle; tumour-related factors include histological subtype, aggressiveness of the tumour, and spread of disease at diagnosis; and, healthcare system factors include the availability and quality of early detection, diagnostic and treatment services. Examined across cancer types and regions, survival estimates can be used to establish priority areas for improving prognosis.Note 1 Examined over time, and in conjunction with incidence and mortality trends, survival estimates can be used to monitor progress in cancer control.Note 2 Because of the importance of cancer survival, the Canadian Cancer Registry (CCR) regularly produces survival estimates using the cohort approach. Key aspects of the methodology employed are detailed below.
Survival analyses include all primary cancers, including multiple primaries for the same person. This approach is becoming standard practice.Note 3-5 However, cancers diagnosed through autopsy only or death certificate only (DCO) are excluded from survival analysis because the date of diagnosis, and thus survival time, is unknown. Since the “true” survival of cases registered as DCO is generally poorer than those registered by other means, the common approach of excluding DCOs may bias survival estimates upward, particularly in provinces/territories with proportionally more DCOs. The magnitude of such bias, however, is generally minor.Note 6
The vital status of a person with cancer is determined through linkage with the Canadian Vital Statistics Death Database and information reported by provincial/territorial cancer registries (PTCRs). Deaths reported by PTCRs but not confirmed by linkage are included in survival analyses using the date of death submitted by PTCRs. Survival time is calculated as the number of days between the date of diagnosis and date of death or date of last follow-up (whichever is earliest). For the small percentage of persons missing month and/or day of diagnosis or death, the survival time is estimatedNote 7; however, decedents with an unknown year of death are excluded from survival analyses.
Survival analyses are performed using publicly available SAS programs to which minor adaptations are made.Note 8 The standard five-year observation time for each individual is split into multiple observations, one for each interval of follow-up time. Three month intervals are used for the first year of follow-up and six month intervals for the remaining four years for a total of 12 intervals. Since the employed actuarial life table method assumes deaths are evenly distributed within an interval, more intervals are used in the first year of follow-up because mortality is often highest and most unevenly distributed during the first year after a cancer diagnosis. With the exception of cases previously excluded because they were diagnosed through autopsy only or DCO, persons with the same date of diagnosis and death are assigned one day of survival because the SAS program automatically excludes cases with zero days of survival. Survival estimates are then calculated at discrete points in follow-up by taking the product of the interval-specific (conditional) survival estimates within the follow-up period.
Expected survival proportions are derived from sex- and province/territory-specific annual life tables by applying the Ederer II approach.Note 9 Due to small populations, only abridged life tables are produced for Prince Edward Island and the three territories. Using methods suggested by Dickman et al.Note 10, abridged life tables are expanded to complete life tables using the abridged and complete life tables for Canada. Since abridged life tables only extend to age 99 years, expected survival proportions for age 100 to 109 years are drawn from complete Canadian life tables.
Five-year observed survival is the percentage of people surviving five-years after cancer diagnosis. Five-year relative survival ratios are estimated as the ratio of the observed survival of the group diagnosed with cancer to the expected survival for the corresponding general population of the same age, sex, province of residence, and time period. In theory, relative survival ratios greater than 100% indicate that the observed survival of people with cancer is better than that expected in a comparable group from the general population. In these instances, it could be that the persons diagnosed with cancer experienced lower mortality from other causes of death because of a greater than usual interaction with the healthcare system. However, estimates of relative survival greater than 100% should be interpreted with caution as several other factors may be at play including random variation in the observed number of deaths, failure to register some cancer deaths, and imprecision in the estimation of expected survival.
As an indication of the level of statistical uncertainty in survival estimates, confidence intervals formed from standard errors estimated using Greenwood's methodNote 11 are provided. To avoid implausible lower limits less than zero and/or upper limits greater than one for observed survival estimates, asymmetric confidence intervals based on the log (-log) transformation are constructed. Relative survival ratio confidence limits are then derived by dividing the observed survival limits by the corresponding expected survival proportion.
Because survival estimates vary with age and the age distribution of cancer cases can vary over time and between geographic areas, it is usually preferable to use age-standardized survival estimates to compare survival over time, across provinces, or between a province and Canada as a whole. Age-standardized survival estimates are interpretable as the survival estimate that would have occurred if the age distribution of the cancer group under study had been the same as that of the standard population. Age-standardized estimates are calculated using the direct method. Specifically, age-specific survival estimates for a given cancer are weighted to the age distribution of persons diagnosed with that cancer over a recent, relatively long period with the age categories used in the weighting being dependent on the cancer under study. Such an approach has the advantage of producing age-standardized survival estimates that are similar to non-standardized estimates.Note 12 Specifics regarding the standard population used and age categories employed are generally detailed in the various publications released by Statistics Canada. Confidence intervals for age-standardized relative survival ratios are formed by multiplying the corresponding age-standardized observed survival lower and upper limits by the ratio of the age-standardized relative survival ratio to the age-standardized observed survival.